Moonshot AI’s Kimi K2 outperforms GPT-4 in key benchmarks — and it’s free
venturebeat.com
Chinese AI startup Moonshot releases open-source Kimi K2 model that outperforms OpenAI and Anthropic on coding tasks with breakthrough agentic capabilities and competitive pricing.
  • @[email protected]English
    4•14 days ago

    Based FOSS but id take the claims of coding ability with a entire salt mine, LLMs simply arent all that great in general at coding no matter what tech Bros try to tell you.

    • ☆ Yσɠƚԋσʂ ☆OP
      1•14 days ago

      It depends on the task and the specific LLM. My experience is that they can do a lot of things effectively nowadays, and they’re improving rapidly.

      • @[email protected]English
        2•14 days ago

        They really cant, they’re good for boilerplate but thats about it. Every example of a AI heavy code base I’ve seen has been poorly optimised security hole ridden garbage.

        • ☆ Yσɠƚԋσʂ ☆OP
          1•14 days ago

          I can tell you for a fact that they can. However, even managing boilerplate and repetitive code is a huge benefit. Furthermore, these tools are great at combing through code bases and helping you find where you need to make changes in code. If you haven’t actually used these tools in a real project yourself then you don’t really know what they’re capable of.

  • The Infinite NematodeEnglish
    3•14 days ago

    The best implementation of LLM-AI for me is self hosted but accessible on my phone so it’s mine and under my control with my data scoping.

    Thinking like Immich, for LLM-AI.

    Sounds like this steps us closer which is great.